Makkoen/whisper-large-v3-cit-do01-wd0-lr3e-06-FULL5

./7326

This model is a fine-tuned version of openai/whisper-large-v3 on the 7326 FULL-2024-10-24 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3911
• Wer Ortho: 22.6474
• Wer: 15.5576

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-06
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 300
• training_steps: 1600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.686	0.4851	200	0.4602	26.0150	18.7885
0.5255	0.9703	400	0.4216	24.3312	17.1358
0.4328	1.4554	600	0.4028	23.2291	15.9895
0.4064	1.9406	800	0.3945	23.2291	16.1897
0.3579	2.4257	1000	0.3945	22.8195	15.7618
0.3409	2.9109	1200	0.3894	22.6884	15.5812
0.3131	3.3960	1400	0.3909	22.6556	15.6008
0.3021	3.8811	1600	0.3911	22.6474	15.5576

Framework versions

• Transformers 4.45.1
• Pytorch 1.13.1+cu117
• Datasets 3.0.1
• Tokenizers 0.20.0